Educational Psychology in Latin America: With Linear Hierarchical Models Jesús Silva 1(&) , Darwin Solano 2 , Claudia Fernández 2 , Ligia Romero 2 , Nataly Orellano Llinás 3 , Ana María Negrete Sepúlveda 4 , Luz Estela Leon Coronado 3 , and Rosio Barrios González 3 1 Universidad Peruana de Ciencias Aplicadas, Lima, Peru jesussilvaUPC@gmail.com 2 Universidad de la Costa, St. 58 #66, Barranquilla, Atlántico, Colombia {dsolano1,cfernand10,lromero11}@cuc.edu.co 3 Corporación Universitaria Minuto de Dios – UNIMINUTO, Barranquilla, Colombia nataly.orellano@uniminuto.edu, luzleoncoronado@gmail.com, rosiobarriosgo@hotmail.com 4 Universidad Cooperativa de Colombia campus Montería, Montería, Colombia ana.negrette@campusucc.edu.co Abstract. Research in clinical psychology, since its inception, has been aimed at analyzing, predicting and explaining the effect of treatments, by studying the change of patients in the course of them. To study the effects of therapy, research based on quantitative analysis models has historically used classical methods of parametric statistics, such as Pearson correlations, least squares regressions Student’s T-Tests and Variance Analysis (ANOVA). Hierarchical linear models (HLMs) represent a fundamental statistical strategy for research in psychotherapy, as they allow to overcome dependence on the observations usually presented in your data. The objective of this work is to present a guide to understanding, applying and reporting HLMs to study the effects of psychotherapy. Keywords: Hierarchical linear models Á Growth curve models Á Multilevel models Á Psychotherapy 1 Introduction The data used in clinical psychology research is usually nested. This means that they are grouped into hierarchical structures that have different levels, which would make the presence of correlations between observations belonging to the same grouping level [1] expected. To illustrate the phenomenon of data nesting that usually occurs in clinical psychology, let’s imagine a study aimed at analyzing change levels in a sample of 100 patients who were exposed to 20 sessions of a treatment and who completed a measure of clinical severity, at the beginning of therapy (t1), mid-treatment (t2) and at the end of 20 sessions (t3) [2]. In this case, it is expected that a patient’s scores at the time t1 correlate with their own scores at the time t2 and t3; In this way, we will say © Springer Nature Singapore Pte Ltd. 2019 G. Wang et al. (Eds.): iSCI 2019, CCIS 1122, pp. 233–242, 2019. https://doi.org/10.1007/978-981-15-1301-5_19 aviloria7@cuc.edu.co